F5: How to Improve AI Efficiency Further: New Devices, Architectures and Algorithms
2022 IEEE International Solid- State Circuits Conference (ISSCC) · 2022
Machine learning (ML) algorithms and applications continue to evolve at a rapid pace relative to Moore's Law. There is simultaneously a demand for bigger and more complex ML models, ever-growing computational throughput and improved energy efficiency over the coming decade. As we start to hit the limits of technology scaling, what are the latest design strategies to improve performance and energy efficiency of machine learning processors of the future? Further, can ML-based tools improve hardware design methodologies? This forum aims to explore novel circuits, architectures, algorithms, as well as ML-based chip design tools that will push the limits of AI efficiency.